You are on page 1of 27

STRUCTURAL CONTROL AND HEALTH MONITORING

Struct. Control Health Monit. (2014)


Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/stc.1655

Crack detection and characterization techniques—An overview

Yao Yao*,†, Shue-Ting Ellen Tung and Branko Glisic


Department of Civil and Environmental Engineering, Princeton University, Princeton, NJ 08544, USA

SUMMARY

Crack occurrence and propagation are among critical factors that affect the performance and lifespan of civil
infrastructures such as bridges, pipelines, and so on. As a consequence, numerous crack detection and character-
ization techniques have been researched and developed in the past decades in the areas of SHM and non-
destructive evaluation (NDE). The significant amount of performed studies and the large number of publications
give rise to the need to systematize, condensate, and summarize this enormous effort. The aims of this paper
are to summarize the knowledge about cracking and its sources, review both existing and emerging methods
for crack detection and characterization, and identify the advantages and challenges for these methods. In general,
this paper identifies two sensing approaches (direct and indirect) and two data analysis approaches (model-based
and model-free or data-driven) along with a range of associated technologies. The advantages and challenges
of each approach and technology are discussed and summarized, and the future research needs are identified.
This paper is intended to serve as a reference for researchers who are interested in crack detection and
characterization as well as for those who are generally interested in SHM and NDE. Copyright © 2014 John
Wiley & Sons, Ltd.

Received 11 June 2013; Revised 23 January 2014; Accepted 17 February 2014

KEY WORDS: crack detection and characterization; crack typology; structural health monitoring; non-destructive
evaluation; direct sensing; model-based and model-free data analysis

1. INTRODUCTION
Civil structures and infrastructures provide essential welfare in society. Therefore, they are invaluable
assets that need to be preserved and well maintained. In order to achieve this goal, it is necessary to
assess their health conditions. A particular challenge is early-crack detection and characterization
(localization and quantification), because the size of the crack is small at that stage. Cracks can be
defined as unintentional discontinuities in structural materials. In general, they are the consequence
of local material failure. They can be generated by many factors, such as external loads, buckling,
fatigue, physical processes, such as freeze-thaw cycles, and chemical processes, such as alkali–silica
reaction and corrosion. Typically, the cracks that require attention usually initiate with the opening size
of 0.05–0.1 mm (0.002–0.004 in) approximately; they become of true concern at approximately
0.3–0.4 mm (0.013–0.016 in) and can develop to dangerous sizes of several millimeters to several
tens of millimeters. While small cracks may lead to inadequate serviceability, large cracks can result
in structural failures. Early-age cracking in concrete can seriously affect a structure’s durability.
In brittle materials, the occurrence of cracks may lead to immediate failure. However, in ductile
materials, crack growth is usually gradual. It is important to detect all types of cracks before they
reach a critical level and cause the failure of structural components or entire structures. Currently
applied methods for crack detection and characterization include periodic visual inspections and

*Correspondence to: Yao Yao, Department of Civil and Environmental Engineering, Princeton University, E306 EQuad,
Princeton, NJ 08544, USA.

E-mail: yaoyao@princeton.edu

Copyright © 2014 John Wiley & Sons, Ltd.


Y. YAO, S.-T. E. TUNG AND B. GLISIC

non-destructive evaluation (NDE), which is based on traditional instrumentation and approaches [1].
Traditional methods can be expensive, time consuming, and frequently inaccurate during early
stages of cracking by failing to detect damage. Visual inspections can be unreliable because some
types of cracks are sparse or inconspicuous. Furthermore, if a crack occurs between two inspections,
it may propagate and reach a critical size before it can be properly detected and characterized.
Therefore, there is a need for more accurate, reliable, and affordable methods for quasi-real-time crack
detection and characterization. Quasi-real-time assessment can be achieved through continuous
monitoring.
A crack not only changes the local strain field in the structure but also affects other properties such
as electrical conductivity, wave propagation properties, and so on. Cracks can change the global strain-
field distribution, natural frequency, and the shape of the structure. Additionally, the initiation and
propagation of cracks may generate acoustic waves. All these parameters can be, to a certain extent,
used to detect and characterize the cracks. The most frequently applied techniques are based on strain
sensors (electrical or fiber-optic), piezoelectric and acoustic sensors (acoustic emission and wave
propagation monitoring), and accelerometers (modal analysis). Emerging approaches include micro-
electromechanical systems (MEMS), piezoelectric paints, nano-technologies, large-area electronics
(LAE), and laser-based noncontact sensing techniques. In addition, several data analysis algorithms
can be combined with sensing technologies in order to detect and characterize cracks.
This paper is intended to serve as a reference for researchers who are interested in crack detection
and characterization and/or general SHM and NDE methods. The main focus is on civil structures
and infrastructure. It first presents a typology of cracks, including their causes, manifestations, and con-
sequences. The following sections contain summaries of currently available sensing technologies and
data analysis algorithms, identifying the advantages and challenges of each method. The paper ends
with an overview of emerging technologies and conclusions.

2. CRACK CLASSIFICATION, CAUSES, AND CONSEQUENCES


2.1. Crack classification
Concrete (reinforced and prestressed) and steel are the most widely used construction materials in prac-
tice. Because both materials can experience cracking due to various causes, this paper focuses primar-
ily on cracks in concrete and steel. A general method of classifying cracks is to identify them on the
basis of their geometry [2]. On the basis of its direction of propagation, appearance, and position, a
given crack in a beam structure can be classified as follows [2]:
• transverse cracks are perpendicular to the longitudinal axis of the beam structure;
• longitudinal cracks are parallel to the longitudinal axis of the beam structure;
• slant cracks have an angle with respect to the longitudinal axis of the beam structure;
• breathing cracks open when subjected to tensile stress and close when the stress is reversed;
• gaping cracks, also known as notches, remain open and do not close;
• surface cracks occur on the surface of the structure and can commonly be detected by visual
inspection (except for buried, coated, or remote structures) or surface-mounted sensors; and
• Subsurface cracks propagate below the surface of the structure and can be detected by acoustic
emission or wave propagation monitoring methods.
Transverse cracks are the most common and most dangerous cracks, because they can reduce a
structure’s cross section and therefore lower its structural capacity. Although longitudinal and slant
cracks are less common, they can affect a structure’s torsional capacity. Transverse breathing cracks
may affect structures in the same manner as regular transverse cracks; that is the reason why many
researchers focus on this type of crack. On the other hand, transverse gaping cracks are the easiest
to simulate under laboratory conditions; therefore, most experimental studies involve gaping cracks.
Research has indicated that with regard to the dynamic behavior of structures, surface cracks are more
important than subsurface cracks [3]. Surface cracks often facilitate adverse chemical leaching and can
accelerate material deterioration, especially in concrete structures.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
CRACK DETECTION AND CHARACTERIZATION TECHNIQUES—AN OVERVIEW

2.2. Cracks caused by mechanical actions


In general, cracks can be caused by mechanical actions (e.g., static, dynamic, and cyclic loadings) or by
physical and chemical processes (e.g., shrinkage, thermal effects, corrosion, alkali–silica reaction, etc.).
Mechanically generated cracks can occur from several causes and vary in different materials. For
concrete and steel, typical mechanisms include the following:
• cracking due to excessive external static or dynamic loadings (e.g., the material reaches critical
levels of stress or strain);
• cracking due to repetitive or cyclic dynamic loading (e.g., fatigue);
• cracking due to stress concentrations (e.g., at the location of abrupt cross-section changes, dents,
grooves, inclusions, forging flaws, material porosities, voids, etc.); and
• cracking due to buckling (e.g., [4]).
A typical sequence of progressive mechanical cracks leading to structural failure is described as
follows [2]:
Crack initiation stage. In this stage, minute discontinuities occur in the material. Cracks can initiate
from any or a combination of the aforementioned mechanisms. During the initiation stage, crack open-
ings typically remain below 100 μm (0.004 in) in width and are thus, in the most cases, invisible to the
human eye.
Crack propagation stage. In this stage, the initial discontinuities grow in size and propagate along a
certain orientation. This is mainly caused by the variation of external loadings and environmental fac-
tors such as increased stresses, fatigue due to cyclic stresses, and changes in temperature and moisture.
With respect to the structural capacity of reinforced concrete, crack widths of concern are usually about
0.33 mm (0.013 in) for internal exposure and 0.41 mm (0.016 in) for external exposure [5]. Nuclear
reactors or waste treatment plants, however, have stricter regulations on crack size control. Conditions
that can accelerate the crack growth rate include residual stresses, thermal stresses, metallurgical con-
ditions such as hydrogen in steel and carbide precipitation in alloy steels, and environmental conditions
such as corrosive mediums.
Failure stage. When the size of a crack exceeds a certain size, the structure will no longer be able to
support the applied loads and will consequently fail. Once the crack reaches its critical length, the
failure process is rapid.
The most common cases of mechanically generated cracks are presented in more detail in the
succeeding text.
Concrete structures are particularly susceptible to cracks from external loadings because they
usually start developing small structural cracks at loads well below service level [6]. Two main types
of structural cracks are flexural cracks and shear cracks, schematically shown in Figure 1. Flexural
cracks are frequently transversal and found in locations subjected to large moments. On the other hand,
shear cracks occur in areas with dominant shear and are identified by their slant (diagonal) direction [6].

Figure 1. Left: a flexural (transverse) crack in concrete; right: a shear (slant) crack in a concrete beam.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
Y. YAO, S.-T. E. TUNG AND B. GLISIC

While flexural failures are usually ductile because of the presence of steel reinforcement, shear failures
tend to occur suddenly with no advance warning [6]. Therefore, shear cracks are generally considered
more detrimental for reinforced concrete structures.
Concrete structures can also develop fatigue cracks. While the fatigue mechanism in concrete is not
clearly understood because of the material’s heterogeneous composition, studies have shown that
fatigue loading on concrete initiates the growth of microcracks at the locations of internal flaws
[7,8]. Fatigue crack propagation in concrete depends on the interaction between microcrack growth,
which encourages further cracking, and aggregate crack bridging, which prevents further cracking
[8]. Depending on the stress magnitude and direction, concrete structures can take more than one
million cycles to fail. However, fatigue cracks can cause excessive deformations, large crack widths,
and reinforcement debonding—all of which can lead to severe structural failures [9].
Steel structures can exhibit cracking if a critical level of loading or the material’s yield strength
is reached. Depending on material composition and treatment, the yield stresses of structural steel
range from 220 MPa (32 ksi) for carbon steel up to 690 MPa (100 ksi) for quenched and tempered
steel [10]. Steel structures may develop brittle and/or ductile cracks if they undergo buckling or
torsion from external loads. These cracks develop at locations of maximum tension and compression
and propagate perpendicular to the direction of the stresses [11]. Ductile cracks exhibit plastic
deformation at their edges and are formed by the coalescence of microvoids that develop from
inclusions [11]. Gradual failure occurs when the voids reach a critical size and form a central crack.
In contrast, brittle cracks in steel show very little plastic deformation and develop similarly to concrete
cracks [11]. Because brittle failure occurs suddenly and rapidly, it is therefore more critical than ductile
failure.
Fatigue cracks (Figure 2) occur in steel because of repeated loading cycles. They usually originate
on the steel surface because of some discontinuity or stress concentration [11]. After initiation, the
crack propagates perpendicular to the direction of maximum tension and compression [12]. The load-
ing opens and blunts the crack tip and the unloading re-sharpens the tip. This repeated process is
accompanied by shear, which causes further crack growth [13]. Studies have determined linear and
log-linear relationships between the surface crack length and the crack depth [14–17]. While the exact
aspect ratio of crack depth to crack length depends on the specific material properties, the shape of a
crack front in the interior of the steel normally has a semi-elliptical shape [15]. Fatigue failure accounts
for over 80% of all failures in metals because it occurs at stresses well below the yield strength of the
material [13,18]. While it takes several years for fatigue cracks to grow, once they reach a critical size
the structure will undergo brittle failure with little deformation [11]. Thus, the consequences of fatigue
cracks in steel can be severe.

2.3. Cracks caused by physical and chemical processes


In addition to mechanical cracks, concrete and steel can develop cracks due to various physical and
chemical processes. These processes, which depend on the particular environmental and chemical
properties of the material, typically include the following:

Figure 2. Example of fatigue crack in steel.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
CRACK DETECTION AND CHARACTERIZATION TECHNIQUES—AN OVERVIEW

• Cracking in concrete due to


- early-age deformation (endogenous and plastic shrinkage);
- drying shrinkage;
- freeze-thaw cycles;
- temperature variations and gradients;
- sulfate attack;
- alkali–aggregate reaction; and
- chloride penetration and corrosion in reinforcement bars (rebars)
• Cracking in steel due to stress corrosion.
Shrinkage in hardened concrete can be subdivided into three main categories: drying shrinkage,
plastic shrinkage, and endogenous shrinkage [19]. Drying shrinkage strongly depends on the humidity
in the environment and the water–cement ratio of the concrete [19]. When wet concrete dries, the evap-
oration of water develops suction in the pore liquid, which forces the concrete to contract [20]. Early-
age plastic shrinkage occurs when the rate of water evaporation from the concrete surface exceeds its
bleeding rate, resulting in high capillary stresses near the surface [21]. Endogenous shrinkage is less
dependent on the environment and refers to the shrinkage caused by chemical processes (carbonation
and hydration) and thermal activity (heat of hydration dissipation) in the concrete [19]. For all cases,
shrinkage is greatest at the surfaces of the structure. This causes tensile stresses that can exceed the
concrete’s tensile strength and develop into surface cracks [19]. Shrinkage cracks are very fine, rarely
exceeding 10 mm (0.4 in) in depth and 10 μm (0.0004 in) in width [20]. They are of varying lengths
and are generally spaced from a few centimeters up to 3 m (10 ft) apart [22]. In general, shrinkage
cracks can accelerate the degradation process of concrete by encouraging waterproofing, chemical
leaching, and corrosion problems [23]. An example of shrinkage crack is shown in Figure 3.
Freeze-thaw concrete cracking occurs at locations prone to water saturation and freezing [24]. Above a
critical degree of saturation (91.7% of water), ice crystals in concrete will exert a hydraulic pressure inside
the pores [25]. If this pressure exceeds the tensile strength of the surrounding material, the structure will
develop subsurface cracks that propagate to the surface. Successive freeze-thaw cycles can cause not only
cracks, but also pop-outs. A pop-out appears as a circular crack and leaves a crater on the concrete surface;
their dimensions range from a few millimeters to 100 mm (4 in) in diameter and up to 40 mm (1.6 in) in
depth [20]. Freeze-thaw damage typically takes several years to appear on the concrete surface so they
are not as critical as other types of cracks. However, if untreated, they can eventually lead to the total
disintegration of concrete [24]. An example of a freeze-thaw crack is given in Figure 4.
Concrete can undergo chemical attacks when certain chemical solutions leach into the structure. The
presence of sulfate facilitates the expansive formation of ettringite or gypsum inside the concrete.
Because the chemical products are more voluminous than the surrounding cement, they will induce
an internal pressure that may be sufficiently high enough to cause parallel subsurface cracking [20].
Extensive cracking inside the structure will eventually lead to the exfoliation and delamination of
concrete surfaces. Additionally, alkali–silica reactions occur when the cement mixture is too alkaline.
Silicate ions in the aggregate can react and form alkali-silicate gel, a product that is also more volumi-
nous than concrete [20]. Isolated cracks form during the early stages of alkali–silicate reaction, termi-
nating at depths of 50–75 mm (2–3 in). However, the cracks form joined networks as the reaction
develops, as shown in Figure 4. This may eventually lead to exfoliation and pop-outs [20]. While they

Figure 3. Shrinkage crack.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
Y. YAO, S.-T. E. TUNG AND B. GLISIC

Figure 4. Left: freeze-thaw crack (source: inspectapedia.com, photo: Daniel Friedman [109]); right: cracks formed by
alkali–silica reaction (source: wikipedia.org, photo: Dr. P.E. Grattan-Bellew, Federal Highway Administration [110]).

may take several years to develop, chemical cracks eventually lead to material loss, which severely
reduces the structure’s capacity and increases its vulnerability to further chemical attacks.
Corrosion cracks can occur in reinforced concrete because of rebar corrosion. When the pH of the
pore solution in concrete is high, a thin oxide layer forms on the surface of reinforcement [20]. If
exposed to low-pH moisture, the protective oxide layer will break down and the steel will react with
water, forming rust. This typically happens when chloride penetration (e.g., dissolved de-icing salt)
reaches the rebars. Because the volume of reaction products is greater than the initial volume of rebars,
an expansive pressure will form inside the concrete and cause rupture [26]. The resulting cracks begin
adjacent to the rebars as subsurface cracks and can propagate to the surface. While they do not have
urgent consequences when they are internal, rebar corrosion cracks become extremely critical once
they cause the surrounding concrete to spall as this reduces the cross-section area of structural members
and further expose rebars to accelerated corrosion [20]. An example of a corrosion crack is shown in
Figure 5.
Temperature gradients and differential thermal expansions can have damaging effects on structures.
Globally, temperature variations may generate stresses that facilitate mechanical cracking in statically
indeterminate structures. Locally, they can change the mechanical properties of the material and lead to
its deterioration, especially in the event of fire. Structural steel elements are usually fireproofed, and
therefore, thermal cracking is more prevalent in concrete. Initial heating evaporates the pore water
and actually causes cement paste contraction. However, the aggregate undergoes progressive
expansion, opposing the shrinkage and possibly rupturing the surrounding cement [20]. Fire damage
can cause random cracking networks and surface spalling, and the latter is more critical because it is
explosive and unpredictable [20]. Heating below 250 °C frequently does not cause stresses large
enough to cause significant cracking [20].
Steel elements may undergo stress corrosion cracking (SCC) when they are simultaneously exposed
to tensile stresses and corrosive environments such as chlorides (Figure 6). In general, low carbon
steels become more susceptible to SCC as the carbon content increases [11]. Studies have shown that
there is a threshold tensile stress intensity, below which SCC does not occur. This stress intensity
varies depending on the metal, heat treatment, and surrounding environment [11]. SCC cracks usually
form at physical discontinuities due to chemical accumulations and propagate along the surfaces of
bends and welds due to high residual stresses. Initially, the cracks are very fine and penetrate deeply

Figure 5. Reinforcement corrosion crack.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
CRACK DETECTION AND CHARACTERIZATION TECHNIQUES—AN OVERVIEW

Figure 6. Stress corrosion cracking in steel (source: Metallurgical Technologies, Inc. [111]).

into the material without obvious signs of deterioration. However, once the cracks grow to a critical
size, brittle fracture can suddenly occur [11].

2.4. Summary on crack classification, causes, and consequences


The various types of cracks found in the two most common materials in civil engineering practice are
described in this section and concisely presented in Tables I and II. In summary, cracks have adverse
effects on structures; initial cracks may be caused by a single mechanism, but once it has been created,
it can trigger or accelerate other processes that could cause crack propagation. For example, a mechan-
ical crack may act as a channel for accelerated penetration of chemicals, which can then in turn facil-
itate rebar corrosion or alkali–silicate reaction. The delamination of concrete due to rebar corrosion
reduces the cross-section capacity of the structure, which can lead to further crack initiation or growth.
Consequently, all cracks can potentially cause a dramatic decrease in structural capacity and durability.
It is therefore important to detect cracks during their early stages so that any maintenance and preser-
vation work can be planned timely and efficiently.

3. CRACK DETECTION AND CHARACTERIZATION TECHNIQUES BASED ON


DIRECT SENSING
Sensor-based crack detection has been proven effective when the crack occurs close to or in direct con-
tact with the sensors (e.g., [27,28]) or sensing medium (e.g., elastic waves). The crack is detected and
localized directly as an unusual change in the output from the sensors affected by the crack. Hence, this
approach in crack detection and localization is called direct sensing in this paper. Typical examples of
technologies that enable direct sensing include discrete strain sensors (electrical or fiber optic, short-
gage or long-gage), distributed strain sensors (electrical or fiber optic), wave monitoring sensors
(acoustic emission and wave propagation sensors), and eddy current sensors. Discrete and distributed
strain sensors have to be in direct contact with the potential cracks in order to detect them, while wave
monitoring and eddy current sensors can detect cracks as a perturbation in the elastic wave or electric
field patterns. These techniques are described as follows.

3.1. Discrete strain monitoring


A material fails when the stress at a certain point exceeds the strength of material. Strain is a parameter
directly correlated to stress. There is no effective means to monitor stress under real, on-site conditions;
consequently, strain is one of the most relevant parameters to monitor [29]. Electric and fiber-optic
sensors are the two most used strain-monitoring technologies. Each technology features discrete and
distributed sensors.
Resistive strain gages (see examples in Figure 7) are the most mature and widely used sensors for
strain monitoring. They are commercially available in many forms, with gage lengths ranging between
0.4 and 100 mm (0.016 and 4 in). The strain gages are usually bonded to the surface of the structure by
suitable adhesives. A strain change in the structure would be transmitted to sensor, which could then be
accurately measured by an appropriate reading unit. Strain gages have a long history in crack detection.
For example, Heise [30] presented a laboratory test on a steel box-like structure in which the strain

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
Copyright © 2014 John Wiley & Sons, Ltd.
Table I. Steel crack summary.
Mechanical causes Chemical causes
Loading Fatigue Stress corrosion
Causes Loading stresses that exceed the yield Repeated loading cycles (especially at locations Simultaneous exposure to tensile stresses and
strength of discontinuities or stress concentrations) corrosive environments
Physical appearance Perpendicular to the direction of maximum Perpendicular to the direction of maximum Form at physical discontinuities and propagate
principal stress principal stress along the surfaces featuring high residual stresses
Size and shape Crack size and shape vary depending on loading Initial cracks are minute. Lengths and depths are Initial cracks are minute but penetrate deeply into
correlated; crack fronts are usually semi-elliptical the material
Consequences Ductile failure is gradual. Brittle failure is sudden Brittle failure at stresses well below the yield strength Brittle failure (sudden and rapid)
and rapid
Y. YAO, S.-T. E. TUNG AND B. GLISIC

DOI: 10.1002/stc
Struct. Control Health Monit. (2014)
Table II. Reinforced concrete crack summary.
Mechanical causes Physical and chemical causes
1 2
Loading and fatigue Shrinkage , freeze-thaw , and thermal3 Chemical reactions and corrosion
1
Causes Bending moments cause flexural Evaporation causes drying and plastic shrinkage, Leaching of sulfate solutions and the presence

Copyright © 2014 John Wiley & Sons, Ltd.


(transverse) cracks hydration and carbonation cause endogenous shrinkage of contaminated aggregate prone to alkali–
silicate reaction
2
Shear stresses cause shear (slant) Water saturation (>91.7%) and freezing temperatures Reinforcement bars exposed to penetrated
cracks chlorides corrode and expand
3
Repeated loading cycles cause Temperature gradients and differential material expansions
fatigue cracks
1
Physical appearance Perpendicular to the direction of Shrinkage causes fine surface cracks Alkali–silicate reaction causes surface crack
2
maximum principal stress Freeze-thaw causes surface cracks and pop-outs networks, exfoliation, and pop-outs
3
Temperature causes surface crack networks and spalling Rebar corrosion causes subsurface cracks and
spalling.
1
Size and shape Loading crack widths vary depending Shrinkage cracks vary in length but rarely exceed 10 mm in Initial alkali–silicate reaction cracks are
on loading type depth and 10 μm in width typically 50–75 mm deep
2
Fatigue cracks are initially minute, but Size of freeze-thaw surface cracks vary; pop-outs are usually Size and depth of corrosion cracks can vary
developed cracks may have large widths <100 mm in diameter (depending on the position of rebars and
3
Thermal cracks vary in size concrete layer below surface)
1,2
Consequences Flexural failures are gradual Shrinkage and freeze-thaw cracks facilitate further Loss of concrete and significant reduction
Shear failures are sudden deterioration from freeze-thaw cycles and chemical attacks in strength capacity
3
Fatigue cracks may lead to excessive In addition, thermal cracks may cause concrete loss and a
deformations, reinforcement debonding, significant reduction in strength capacity
and brittle failure
CRACK DETECTION AND CHARACTERIZATION TECHNIQUES—AN OVERVIEW

DOI: 10.1002/stc
Struct. Control Health Monit. (2014)
Y. YAO, S.-T. E. TUNG AND B. GLISIC

Figure 7. Left: an example of resistive strain gage; right: an example of a strain sensor based on vibrating wires
(courtesy of Roctest, Inc.); images are not scaled.

gage was able to follow the initiation and propagation of cracks caused by low-cycle fatigue. This early
test not only proves the sensitivity and accuracy of strain gages in crack detection but also shows the
general ability of strain sensors to detect and characterize cracks. Besides resistive strain gages, other
widely used electrical discrete strain sensors are those based on vibrating wires (Figure 7). Sensors
based on vibrating wires have the advantage of incorporated temperature compensation in the sensor
packaging. They also feature longer gage lengths, typically ranging between 50 and 150 mm (2 and
6 in), and can be both embedded in concrete or mounted on the surface of steel and concrete structures.
The longer gage length makes them more sensitive to crack occurrences because the increased contact
length with the structure increases the probability of the sensor to capture the crack. However, both
resistive strain gages and wire vibrating sensors can be considered as short-gage sensors.
Discrete strain sensors based on fiber-optic technologies have reached market maturity during the
last decade. Their functioning and performance were proven not only in harsh conditions frequently
found on-site [31] but also in very extreme environments [32]. While there are several physical
principles for discrete sensors, the most widely used sensors are those based on Extrinsic Fabry–Perot
Interferometry (EFPI), fiber Bragg grating (FBG), and low-coherence interferometry (SOFO). EFPI
sensors have gage lengths ranging between 51 and 70 mm (2 and 2.75 in), while FBG sensors can
be made with a wider range of gage lengths from 10 mm (4 in) up to 2 m (6 ft 6.7 in). SOFO sensors
are exclusively long-gage sensors with lengths ranging between 250 mm (10 in) up to 20 m (66 ft).
More information on the functional principles of these sensors can be found in literature [31]. Several
field application examples of the aforementioned sensors are also presented in literature [27,31,33,34],
including crack detection in bridge deck, tunnel vaults, and pile foundations. The robustness of these
sensors in damage detection is based on the direct detection principle and outlier identification algo-
rithm in which the signal generated by damage dominates the environmental noise (see further text).
Compared with electric strain sensors, short-gage fiber-optic sensors such as EFPI and FBG (Figure 8)
are more sensitive and durable (they are insensitive to electromagnetic influences, humidity, and

Figure 8. Examples of fiber-optic strain sensors; left: short-gage sensors (courtesy of Roctest, Inc., Fibersensing,
and Micron Optics); right: long-gage sensors (courtesy of Micron Optics and SMARTEC); images are not scaled.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
CRACK DETECTION AND CHARACTERIZATION TECHNIQUES—AN OVERVIEW

corrosion) but have substantially higher cost. Similar to vibrating wire sensors, short-gage fiber-optic
sensors can be both embedded in concrete or mounted on the surface of steel and concrete structures.
The key challenges of short-gage sensors (electrical or fiber optic) are (1) their insensitivity to
strain-field anomalies that do not occur in the immediate proximity of the sensor [35] and (2) the large
number of sensors required to cover a large area, thereby imposing complex cabling, deployment, and
data acquisition (e.g., [30]).
Long-gage fiber-optic strain sensors (e.g., [27,31], Figure 8) enable larger coverage for crack detec-
tion than short-gage sensors because of their longer gage length. Several examples from the past have
shown successful crack detection and characterization (localization and quantification) by long-gage
sensors (e.g., [27,36,37]). Two examples of field applications are given in Figure 9 [27,36]: the crack
crosses the locations of sensors and generates a significant strain increase for both sensors, revealing that
a crack has occurred and indicating location and the size of the crack (more details are given in [36,27]).
Figure 9 shows that direct sensing provides very reliable crack detection, given that the crack is in
contact with the sensor (e.g., crack detection would be successful even if the noise generated by the
variable environmental conditions and the monitoring system were an order of magnitude larger).
However, the reliability of detecting and characterizing damage that occurs in locations far from the
sensors is challenging. In some cases, this problem can be solved by using sophisticated algorithms
(indirect sensing, see the next section). However, the performance of such algorithms may be chal-
lenged in on-site conditions because of various influences that may ‘mask’ the damage such as high
temperature variations and load changes, and outliers and missing data in monitoring results [38].
Although long-gage sensors have improved capabilities for crack detection and characterization when
compared with short-gage sensors, they can also suffer from insufficient spatial sensor coverage.
Figure 10 compares damage detection capabilities of short-gage, long-gage, and distributed sensors
on the basis of their spatial coverage in the structure.

Figure 9. Examples of crack detection and localization in real structures using long-gage fiber-optic strain sensors;
left: in a tunnel [36]; right: in a bridge [27].

Figure 10. Schematic comparison between damage detection capabilities of short-gage, long-gage, and distrib-
uted sensors.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
Y. YAO, S.-T. E. TUNG AND B. GLISIC

The discrete strain sensors (short-gage or long-gage) are usually placed at locations where high
strain is expected and the crack is more likely to occur (e.g., damage ‘A’ is expected in the middle
of a simply supported beam; see Figure 10). However, because of local variability in material proper-
ties and dimensions, the cracks may not necessarily occur exactly at the locations of the highest stress
but at locations that are relatively distant from these locations (e.g., damage ‘B’ or damage ‘C’ occurs
instead of damage ‘A’ in Figure 10). Short-gage sensors would not be able to directly detect the dam-
age in that case as they are not in direct contact with the damage. The long-gage sensors extend over
larger areas, and thus, they are more likely to directly detect the damage that does not occur exactly at
the expected locations (i.e., they can directly detect damage ‘B’), but they are still inefficient if the
damage does not occur at the location of the sensor (damage ‘C’).
In conclusion, discrete sensors (electric or fiber optic, short-gage or long-gage) that are sparsely
spaced in real structures may suffer from insufficient spatial coverage. However, if the approximate
locations of crack occurrences are known (e.g., on the basis of numerical modeling), then long-gage
sensors can provide much better crack detection capabilities than short-gage sensors. Increasing the
number of sensors may help as well but only to a certain extent because this would involve increased
hardware and installation cost.

3.2. Distributed strain monitoring


Distributed strain sensing can overcome the spatial coverage and specificity challenges related to
discrete sensing. A distributed sensor can be represented by a single cable that is sensitive at every
point along its length. Hence, one distributed sensor can replace a large number of discrete sensors.
Moreover, it requires a single connection cable to transmit information to the reading unit instead of
using a large number of connecting cables as in the case of wired discrete sensors. Finally, distributed
sensors are less difficult and more economical to install and operate on large structures. An illustrative
comparison between a pipeline structure equipped with distributed and discrete sensors is shown in
Figure 11 [28].
Coaxial cable sensors [1], as shown in Figure 12, are distributed sensors with a high potential for
detecting cracks and measuring strain in reinforced concrete structures. They are based on electrical
time-domain reflectometry (ETDR). The components of coaxial cable sensors are shown in Figure 12
[1]. The inner conductor is separated from a steel spiral by using Teflon, and the assembly is covered
with a thin layer of solder to create a continuous outer conductor [39]. The sensor is designed to be
embedded in concrete. Cracks developed in concrete structures can lead to the separation of the local
steel spiral, and this large discontinuity can be then detected with an ETDR reading unit. As a result, a
cable sensor in the concrete can detect both the location and width of cracks by calculating the reflec-
tion coefficient, which indicates the intensity of reflected waves through multiple cracks [1]. Chen et al.
[5] performed both laboratory testing and field application of crack detection in a reinforced concrete
bridge in Missouri, and they concluded that the measured reflection coefficient correlates well with

Figure 11. Distributed versus discrete monitoring [28].

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
CRACK DETECTION AND CHARACTERIZATION TECHNIQUES—AN OVERVIEW

Figure 12. Distributed sensors; left: example of coaxial cable [5]; right: examples of fiber optic distributed sensors
(courtesy of SMARTEC); images are not scaled.

crack width. The major challenge for this sensing technique is related to signal losses over large
distances—the site application was based on a 2.3-m-long (90 in) sensor, and the longest measurable
length of the sensor was not documented. Other challenges include the application of the sensor to
existing concrete and steel structures and repair and replacement of the malfunctioning sensor.
There are two main principles for distributed strain sensing in the domain of Fiber optic sensors (FOS)
the Rayleigh scattering effect (e.g., [40]) and the Brillouin scattering effect (e.g., [41]), which can be
stimulated (BOTDA) or spontaneous (BOTDR). Each technique is based on the relation between the
measurand (i.e., strain) and the encoding parameter (i.e., the change in optical properties of the scattered
light). Rayleigh-based sensors feature better spatial resolution (10 mm) than Brilouin-based sensors
(500–5000 mm), but they are limited to modest lengths of 70 m. Stimulated Brillouin scattering is the least
sensitive to optical losses and allows long distance monitoring of up to 400 m without channel switches.
With channel switches, the monitoring distance is practically unlimited [42] (an application involving
application of 5 km on a real steel bridge is given in [41]). The sensors can be both installed on the surface
of existing structures (e.g., [28,41]) and also embedded in the concrete of new structures (e.g., [27]).
Distributed fiber-optic sensing systems based on Brillouin scattering were designed for detecting
cracks with widths larger than 0.5 mm [43]. Examples of crack detection in real on-site conditions
and close-to-real conditions are given in Figure 13.
Similar to discrete sensors, the crack is detected directly as an unusually high strain change. While
the localization of the crack is very simple (position of strain change along the cable), the size of the
crack is challenging to determine because of the averaging of the measurements and the accuracy of
the sensing system. Figure 13 (right) shows the comparison between distributed and long-gage sensors
at the location of the crack opening. The discrete sensors feature better measurement properties (higher
resolution, accuracy, and frequency of reading). However, distributed sensors feature better spatial
coverage.
The packaging of optical fibers in a sensing cable and the surface installation remain as challenges
for distributed fiber-optic sensors, as the former issue introduces optical losses that result in shorter

Figure 13. Crack detection and localization using distributed fiber-optic sensors; left: on a large-scale test of
pipeline exposed to permanent ground movement [28]; right: on a pedestrian bridge due to thermal gradients at
early age of concrete [27].

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
Y. YAO, S.-T. E. TUNG AND B. GLISIC

monitored lengths, and the latter issue is time consuming and cost sensitive. Another important tech-
nical challenge for distributed fiber-optic sensors is repair and replacement of malfunctioning sensor.

3.3. Direct sensing based on wave monitoring


Acoustic emissions occur when discontinuities in the structure such as crack openings release energy.
This energy travels in the form of high-frequency acoustic (elastic) waves, which can provide compre-
hensive information about the origination of a discontinuity (crack) in a stressed component and the
development of this discontinuity (e.g., change in size) caused by continuous or repetitive stresses.
Because acoustic waves travel through the structure, it is possible to localize the position of a crack
by placing various transducers over the structure and determining the wave speed and the propagation
time. Even relatively weak acoustic emissions can be used for crack detection in metallic struc-
tures [44]. Grosse [45] used acoustic transducer network to detect the damage on a prestressed
reinforced concrete bridge built in a large test facility in Germany. The transducers were optimized
to operate on signals up to 25 kHz, and they successfully detected and characterized (localized and
quantified) the crack with the signal processing techniques of beamforming and velocity spectral anal-
ysis. An example of the acoustic transducer is shown in Figure 14. The challenge of this method is in
the creation of complex algorithms that have to guarantee robustness in real on-site conditions, where
many sources of noise are present. Another challenge is the detection of cracks that may occur while
the monitoring system is off.
Piezoelectric transducers (e.g., [46]) can be used as both actuators that generate elastic waves and
detectors of the waves incoming from other piezoelectric elements. Typically, they can introduce elas-
tic waves into the structural member. The propagation properties of waves (such as attenuation, veloc-
ity, or reflections) can be correlated with the health condition of the member, and if a crack is present, it
can be detected and characterized. There are several types of elastic waves that can be used for damage
detection and characterization purposes, and the most frequently used ones are briefly presented here.
P-waves are longitudinal waves propagating through continuous materials. They are commonly used
for the NDE of concrete [47]. Rayleigh waves propagate near the surface of solids, and they are used
to detect defects in steel and concrete [48]. Lamb wave is an elastic wave propagating along a thin solid
(plate), whose particle motion lies in the plane defined with the direction of wave propagation and the
direction normal to the surface of the thin propagating solid (plate). In steel structures, the Lamb waves
can be directed [49,50] and made to propagate over large distances (e.g., in pipelines up to 1 km).
Figure 14 shows an example of the piezoelectric transducer.
By using active sensors (that serve as wave emitters and receivers) and performing appropriate data
analyses, it is possible to detect and localize cracks. The transducers must be carefully bonded to the
structure, and emitter/receiver distance limitations imply the need for a large number of sensors and
cables to cover large areas of the structure. This significantly increases the cost of monitoring. The data
required from all of the sensor channels can also become extremely complex to analyze.
Eddy-current-based crack detection techniques exploit the fact that damage (e.g. cracking) alters the
flow of the eddy current generated in the structure, and this is detected as a variation in the remote mag-
netic field (Figure 15 [51]). The technique is mostly applied on metallic structures that are good electrical
conductors. A driving coil is used to generate the eddy current in the structure while the sensing coil

Figure 14. Left: acoustic transducer; right: piezoelectric transducer (source: model LDT1-028 K, Measurement
Specialties, Inc. [112]); images are not scaled.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
CRACK DETECTION AND CHARACTERIZATION TECHNIQUES—AN OVERVIEW

Figure 15. Top view (left) and side view (right) of the eddy current sensor [51].

is used to detect anomalies in remote magnetic field. Sadler and Ahn [51] presented an integrated eddy
current sensor for proximity sensing and detecting microcracks on metallic surfaces. They demon-
strated that their sensor is capable of detecting crack depths that are as small as 200 μm. Najafi [52]
presented the use of eddy-current-based techniques for damage detection in pipelines. The technique
is typically used in NDE as a movable proximity sensor and therefore does not allow for continuous
monitoring unless numerous transducers are installed on the structure, which is similar to the chal-
lenges in acoustic emission and wave propagation techniques.

3.4. Summary on direct sensing techniques


In summary, direct sensing techniques are successful in crack detection and characterization. Although
they have various advantages, they also face some challenges. Both advantages and challenges are
presented in Table III.

4. CRACK MONITORING TECHNIQUES BASED ON INDIRECT SENSING


(MODELS AND ALGORITHMS)
4.1. General observations
Indirect sensing methods are based on measurements made by sensors that are not necessarily in direct
contact with damage (e.g., accelerometers, strain sensors that are not at the location of the damage, re-
mote sensors, etc.). As opposed to direct sensing where the damage is detected directly as a noticeable
change in output signals of affected sensors, in the case of indirect sensing, the sensors usually do not
manifest noticeable changes in output signals. Consequently, the recorded data have to be analyzed
using various classes of sophisticated algorithms in order to ascertain crack detection and perform
crack characterization. The deployed sensing technique can be based on various types of sensors such
as strain sensors (not in contact with the damage), remote sensors (scanning laser vibrometer), and so
on. However, the techniques that are frequently applied at present are related to measurements
performed with accelerometers. This overview mostly (but not only) focuses on these techniques.
Generally, two approaches in data analysis are identified: model-based and model-free (data-driven)
approaches.

4.2. Model-based approaches


Model-based algorithms use mathematical models describing structural behavior to infer crack occur-
rences, locations, and sizes. A typical example is modal analysis, where the data collected by

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
Table III. Comparison of various direct sensing techniques currently used for crack detection and characterization.
Discrete strain monitoring Distributed strain monitoring Wave-based monitoring
Type of Resistive Electrical vibrating FOS short FOS long Coaxial Eddy
sensor strain gage wires gage gage cable FOS distributed Acoustic Piezoelectric current
Function Electric resistance Vibrating wires EFPI, FBG FBG, SOFO ETDR Rayleigh and Acoustic Wave Eddy
principle change Brillouin emission propagation current
Scattering
Structural All, new and All, new and All, new and All, new and Concrete, new All, new and All, new All, new and Metal, new
material existing structures existing existing existing structures existing and existing existing and existing
and type (surface mounted) structures structures structures structures structures structures structures

Copyright © 2014 John Wiley & Sons, Ltd.


Application Local monitoring, more useful if crack location can be predicted; short gage Global monitoring, more useful if crack location cannot be predicted (all materials);
guidance mostly for homogeneous materials (e.g., metals and composite materials in the acoustic and piezoelectric perform better on metals and composite materials but can
direction of reinforcing fibers), long gage for all materials also be used on concrete
Gage 0.4–100 mm 50–150 mm EFPI: 51–70 mm N/A (Limited due Rayleigh: 70 m Depends on material properties, 0.1–1 m
length/ (0.016–4 in) (2–6 in) (2–2.75 in) to signal loss) (200 ft) wave frequency, and (4–40 in)
length FBG: 0.01–2 m Brillouin: 5 km environmental noise; typical
range (4–80 in) (3 mi) 0.2–70 m (8 in–200 ft)
SOFO: 0.25–20 m
(10 in–66 ft)
Advantage Low cost, high High stability High durability, Improved High spatial High spatial High spatial resolution and High
accuracy, simple and accuracy, stability, and spatial resolution resolution, sensitivity sensitivity
data processing simple data accuracy, resolution, durability, and
processing simple data durability, stability, simple
Y. YAO, S.-T. E. TUNG AND B. GLISIC

processing stability, and data processing


accuracy, simple
data processing
Challenge Low spatial Low spatial High cost, High cost, Signal loss, Rayleigh: short- Complex cabling, affected by Real-time
resolution, resolution, low limited spatial short-length length range environmental noise, complex damage
complex cabling, complex spatial resolution range, surface Both: sensor data analysis detection
long-term drift, cabling, resolution installation, packaging, (mostly
affected by affected by sensor repair complicated used as
electromagnetic electromagnetic surface NDT and
interference interference installation, not SHM)
sensor repair

DOI: 10.1002/stc
Struct. Control Health Monit. (2014)
CRACK DETECTION AND CHARACTERIZATION TECHNIQUES—AN OVERVIEW

accelerometers are processed to calculate the natural frequency, mode shapes, damping, and response
to specifically applied excitation. Changes in these parameters are then associated with cracking
through various algorithms.
Bachschmid et al. [53,54] presented a combined model-based method for crack localization and
depth determination. The method was verified in a laboratory test on a steel beam. The crack position
is determined by using a model-based diagnostic approach and a least-squared identification in fre-
quency domain. The crack depth is determined by comparing the static bending moment generated
by self-weight to identify ‘equivalent’ bending moments that simulate cracks. Ray and Kishen [55] in-
troduced an analytical model to estimate the fatigue crack growth in concrete. The fatigue crack growth
rate depends on several parameters, for example, tensile strength, fracture toughness, loading ratio,
structural size, and so on. This model considers the mentioned parameters and combines them with
the concepts of dimensional analysis. They concluded that this approach is able to capture the size ef-
fect in concrete and that laboratory experimental results supported this method. Ray and Kishen [55]
also found that structural size plays a more important role in determining the fatigue crack propagation
than other factors such as loading ratio and initial crack length.
Friswell and Penny [56] made a review paper about crack modeling in beam structures. They sum-
marize all crack modeling approaches into three main categories: local stiffness reduction, discrete
spring models, and complex models (in two or three dimensions). After numerically comparing differ-
ent methods, they concluded that for low-frequency vibration, local stiffness reduction approaches are
adequate to model the cracks. In addition, they discussed the nonlinear dynamics of a breathing crack
and indicated that with a bilinear stiffness model for crack opening and closing, the impulse and ran-
dom responses of the beam approximate a linear response with natural frequencies. They were able to
accurately detect and localize the cracks in a small beam tested in a laboratory, while the crack depth
estimation would be smaller than reality. Kim and Stubbs [57] presented crack localization and quan-
tification models for beam structures. The models are based on the relationship between fractional
changes in modal energy and changes in a few natural frequencies caused by cracks. The natural fre-
quencies of the tested beams are determined and compared for different crack locations and sizes. The
laboratory tests confirmed that both the crack location and crack size could be estimated with a small
error. Wang and Qiao [58] improved the method by introducing the uniform load surface (ULS) tech-
nique, which combines generalized fractal dimension (GFD) and simplified gapped-smoothing (SGS)
methods. The peak on the GFD curve of ULS indicates both the position and size of the damage, and
the SGS method uses the simple deformation shape of ULS for damage detection. Both methods are
verified through comparisons with laboratory tests involving cracked and delaminated composite
beams.
Dimarogonas and Papadopoulos [59,60] and Papadopoulos and Dimarogonas [61,62] presented a
method based on a flexibility matrix they developed to identify transverse cracks and the influences
of coupling between axial and torsional vibrations and bending on cracks in a beam structure. Miller
et al. [63,64], Brook et al. [65], Rajab et al. [66], and Tsai and Wang [67] also presented methods
for crack localization and quantification based on the analysis of changes in natural frequencies and
mode shapes.
Pai et al. [68] introduced a boundary effect evaluation (BEE) method for localizing cracks and
estimating crack sizes that utilizes operational deflection shapes (ODS) that are measured by a scanning
laser vibrometer. The BEE method extracts boundary solutions from experimental ODS, and cracks are
localized where crack-induced boundary solutions are different from those of actual boundaries. A
local strain energy method is used to estimate the crack size through experimental ODS and stress
intensity factors (from fracture mechanics). Laboratory experiments indicate that the BEE method
can be used to localize and quantify small cracks.
In spite of their abilities to detect and characterize the cracks in laboratory conditions, the aforemen-
tioned methods face several challenges when applied in real-life settings: (1) modal parameters are not
sensitive to very small changes in the strain field when caused by minute cracks (e.g., [69]), which
limits early crack detection; (2) the precise location of the crack is difficult to determine [70] because
all methods suffer from cumulated errors generated by complex data processing; and (3) typical noise
caused by the normal use of the structure (e.g., traffic on the bridge) and environmental factors
(e.g., temperature and humidity variations) may influence the measurements, especially those related

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
Y. YAO, S.-T. E. TUNG AND B. GLISIC

to modal parameters (e.g., [70]). This may result in ambiguous data interpretation. Consequently, the
reliability of crack detection and the accuracy of crack characterization from indirect model-based
approaches are challenged in uncontrolled outdoor conditions, that is, in real-life settings.

4.3. Model-free approaches


Model-free approaches are based on statistical data analysis, which is applied to monitoring data
assuming that a mathematical model of the structural behavior does not exist. These methods involve
the use of sophisticated algorithms such as neural networks [71], fuzzy logic [72], wavelets [73], pattern
recognition [74], robust regression analysis, moving principal components analysis [75], and so on.
Adewusi and Al-Bedoor [76] presented a neural network method for structural crack detection. In
their laboratory experiments, signals of vibration frequencies with and without propagating cracks
were selected to train multilayer feed-forward neutral networks. Other sets of data were tested with
the trained neural networks. They concluded that a simple two-layer feed-forward neural network is
satisfactory in detecting propagating cracks, while trained three-layer networks are successful in
detecting both propagating and non-propagating cracks. Yin et al. [77] proposed a probabilistic
approach to find the crack characterization in a thin-plate structure, which was based on Bayesian
statistical system identification framework. This method only deals with a few points, and the sensors
do not necessarily have to be close to the cracks. The crack characterization is performed on the basis
of statistics, from which the uncertainty and confidence level of crack location, length, and depth are
determined. The authors verified the feasibility of this probabilistic approach with numerical case
studies on a rectangular aluminum plate. Furthermore, Sbarufatti et al. [78] made a combination of
Bayesian Hierarchical Models and artificial neural networks methods to analyze the problem of crack
detection and localization on an aerospace structure. An index output related to crack position and
length is obtained on the basis of the noisy signal inputs, and cracks have been accurately identified
in the majority of the cases.
Umesha et al. [79] presented a method for crack detection based on wavelets. They measured the
deflections of a beam at discrete points and obtained the wavelet coefficients (WCs) to calculate the
position and depth of the crack. They proposed a parametric study by varying the type of damage,
location of damage, intensity of load, flexural rigidity, and length of the beam. The WCs vary because
of the changes of such inputs. In conclusion, they presented a generalized curve to quantify the crack
location and depth in a fixed beam by utilizing the maximum WCs envelope of the deflection response
at the damaged points.
Feldman and Seibold [80] researched an approach that combined model-free and model-based
procedures for detecting the size and location of the damage. First, the Hilbert transform was used
for signal processing, obtaining a slow-varying signal envelope and a phase angle, and detecting
nonlinearities. Then, the Kalman filter was applied for detecting the size and location of the damage
on the basis of the model information provided by the Hilbert transform. Rehman et al. [81] also
presented a method based on the Hilbert transform to detect and classify the nonlinearities in the
flexural and torsional vibrations generated by a crack. Crack depths were determined on the basis of
resonant frequencies and damping levels that are extracted from amplitude-dependent frequency-
response functions.
Vakil-Baghmisheh et al. [82] presented genetic algorithms (GAs) for crack detection in beam-like
structures. They used GAs to detect the changes in natural frequencies of a cantilever beam. The detec-
tion of crack location and depth was transformed into an optimization problem. Both binary and con-
tinuous GAs (BGA and CGA) were utilized to solve the problem by minimizing the optimization
function, which is related to difference between calculated and measured natural frequencies. They
concluded that for BGAs, the average prediction errors of crack location and depth are 1.02% and
1.98%; for CGAs, these errors are 0.73% and 1.11%, respectively. Laboratory experimental results
verified the accuracy of this GA method.
The main challenges for the application of model-free approaches are related to its sensitivity to
noise generated by loading or environmental variations and the lack of supervised training for algo-
rithms. Both challenges can lead to false positive and negative detections of cracks, which reduces
the reliability of such methods in real-life settings.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
CRACK DETECTION AND CHARACTERIZATION TECHNIQUES—AN OVERVIEW

The various types of indirect sensing methods discussed earlier, including model-based testing and
model-free algorithm analysis, are briefly summarized in Table IV. Although this is not an exhaustively
complete list of every method (there are indeed a lot more), this limited selection outlines the diversity of
current research activities.

5. EMERGING MONITORING TECHNOLOGIES AND FUTURE TRENDS


Studies on crack detection and characterization actively and rapidly evolved in the last decade, and
many novel technologies have emerged with the potential to overcome the limitations of the presented
techniques. These emerging technologies include wireless sensors, MEMS, thin-film and piezoelectric
paints, nano-technologies, LAE, and noncontact sensing techniques.
Wireless sensors and sensor networks (e.g., [83,84]) are an emerging sensing paradigm with a great
benefit: they significantly reduce the cost of hardware and installation because external connecting
cables are no longer needed between the sensors and the data acquisition system. In addition, they
allow for denser sensor networks and provide possibilities for distributed data acquisition and analysis
(as opposed to the traditional approach where all the data are collected in one center). Denser arrays of
sensors would improve the direct sensing capability of the monitoring system, while distributed data
management would improve the reliability in data analysis. The current challenges are related to the
power supply of wireless nodes and data losses during wireless transmission. However, ongoing
research in power-harvesting technologies and signal transmission make wireless sensing a very prom-
ising tool for crack detection and characterization.
Micro-electromechanical systems can be equipped with integrated circuits and act as intelligent sen-
sors, actuators, or both (e.g., [85,86]). More importantly, MEMS can be embedded into structures for
subsurface crack detection. Although these lead to sophisticated devices with promising application
potential, these types of possible sensors are somewhat limited. The most common example is accel-
erometers (e.g., [87]), to which modal analysis (having the limitations described previously) is applied.

Table IV. Indirect sensing methods (models and algorithms) for crack detection.
Model-based approaches Model-free approaches
Basic methodology Measure changes of modal parameters Statistical data analysis (pattern
(natural frequency, mode shape, damping, recognition) using sophisticated
etc.) and compare with numerical models algorithms, without comparison
using data analysis methods with mathematical models
Examples of data Model-based diagnostic approach and Neural network method; Bayesian
analysis methods (the frequency domain identification; natural statistical system identification
list is not exhaustive) frequency and mode shape identification; framework; Bayesian hierarchical
static bending moment comparison; model and artificial neural networks
fatigue crack model with dimensional method; wavelets decomposition
analysis, local stiffness reduction, and and wavelet coefficients analysis;
discrete spring model; 2D/3D complex Hilbert transform and Kalman filter;
model, uniform load surface technique genetic algorithms; and so on.
combining generalized fractal dimension
method and simplified gapped-smoothing
method; local flexibility matrix model;
boundary effect evaluation method
utilizing operational deflection shape
measured by scanning laser vibrometer;
and so on.
Advantage Limited number of sensors, contained cost, Limited number of sensors, contained
simple deployment, relatively simple data cost, simple deployment, no need for
analysis numerical modeling
Challenge Low sensitivity to minute crack, low Low sensitivity to minute crack,
accuracy in damage localization, on-site damage localization practically
application (interference with environmental impossible, on-site application
noise) (interference with environmental
noise), complex data analysis

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
Y. YAO, S.-T. E. TUNG AND B. GLISIC

Recently, strain sensors were also developed with MEMS (e.g. [86]). Although low-cost MEMS are
emerging, the need for special processing steps implies increased costs.
Other advanced approaches such as piezoelectric paints (e.g., [88]), carbon nano-tube coatings (e.g.,
[89,90]), and embedded nano-sphere (e.g., [91]) are promising, but these are still not mature enough for
applications to large-scale structures. Sensing skins based on carbon nanotubes are being developed for
2D strain-field monitoring [92]. Sensing skins utilize electrical impedance tomographical conductivity
mapping techniques and provide a 2D mapping of damage [92]. Besides strain-field assessments, sens-
ing skins can be made multi-functional (so that they can detect other defects such as corrosion) [93]
and interfaced using wireless nodes [94]. Another promising approach is the use of multi-functional
materials with self-sensing properties [95]. Recent discoveries point out that the electrical properties
of cementitious materials can be used to detect and locate defects such as cracks [96,97]. Successful
developments and large-scale implementations could practically transform an entire structure into a
sensor.
A novel sensing technology based on LAE [98] can potentially be a good tool for crack detection
and characterization for large-scale structures. Current direct sensing techniques that are based on
strain measurements either monitor cracks at one point or segment (e.g., short-gage and long-gage sen-
sors) or in one dimension (1D, e.g., distributed sensors). LAE sensing sheets can be treated as a quasi-
distributed sensor in two dimensions. The concept of sensing sheet is shown in Figure 16.
Large-area electronics is an emerging technology that allows a broad range of electronic devices to
be integrated on low-cost plastic sheets [99,100]. Through the use of micro-fabrication techniques,
thin-film transducers (including pressure sensors, vapor sensors, particle sensors, etc.) have been
demonstrated, which can be formed into dense arrays spanning large areas (i.e., tens of square meters).
An important benefit of LAE is that it enables the integration of functional thin-film transistors (TFTs).
This means that basic circuit functionality is available to facilitate readout from the large number of
sensor channels. While these TFTs can provide basic functionality, the device-level characteristics that
make them compatible with flexible, large-area substrates also severely limit their energy efficiency.
Thus, large-scale processing over the sensor channels is not viable. Standard electronics technologies
(e.g., based on integrated circuits) have achieved very high efficiency for instrumentation and process-
ing because of nearly five decades of Moore’s-law scaling. Researchers are now exploiting the basic
possible functionality through the TFTs to create specialized interfaces between the large-area sensors
and a potentially large number of readout and processing devices. The individual sensors are read
by embedded integrated circuits with wireless and computational capabilities, taking advantage of
wireless sensing and distributed data management. Finally, the protective layer is made of flexible
photovoltaic, which provides the power needed for the functioning.

Figure 16. Concept of a sensing sheet.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
CRACK DETECTION AND CHARACTERIZATION TECHNIQUES—AN OVERVIEW

Figure 17 shows 2D strain measurements, illustrating how these provide a reliable modality (i.e.,
offering high sensitivity and specificity) for the detection and localization of cracks. In short, LAE
not only can provide a low-cost crack monitoring through a dense and expansive array of sensors
but also can overcome the robustness and reliability limitations affecting current SHM technologies.
In addition, noncontact sensing techniques have also become promising crack detection tools in re-
cent years. Lecompte et al. [101] proposed a crack detection approach with two different camera tech-
niques: a light-emitting diode combined with a charge-coupled device technique (LED-CCD) and the
digital image correlation technique (DIC). Both techniques were used to detect cracks on the surface
of a concrete beam subjected to flexural loading. They measured displacements in several discrete points
on the surface and calculated the deformation using the Green–Lagrange strain expression. It was shown
that these two techniques are complementary: LED-CCD is good for measuring a few data points in a
long time, while DIC performs well for a dense file of data points in a short time. With both techniques,
they confirmed the possibility of detecting the appearance and evolution of cracks even before cracks are
visually observable. Jahanshahi and Masri [102], and Uhl et al. [103] also presented other vision-based
techniques, such as optical instrumentation (e.g., digital cameras), and image processing and computer
vision approaches for non-destructive evaluation of crack damage in real bridge structures.
Liu et al. [104] and Liu and Chen [105] introduced the Terrestrial 3D Light Detection and Ranging
scanner (LiDar) as a remote sensing technique for health monitoring of existing and newly constructed
bridges. They proposed an automated bridge evaluation algorithm called LiDar bridge evaluation
(LiBE) for on-site bridge monitoring, which measured the damage area and operated robustly under
changing environmental conditions. This technique could be applied in crack detection, bridge clear-
ance, and static defection measurement and is especially useful when accurate measurement of bridge
geometry is impossible. However, resolution and complex data analysis are main challenges of this

Figure 17. Illustration of the potential for direct damage detection using high-resolution 2D strain-field measure-
ments. Left-side images show the damage, and right-side images show the expected measurement.

Figure 18. Noncontact laser lock-in thermography (LLT) technique; left: LLT system; right: LLT image for crack
detection in a plate structure [108].

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
Table V. Advantages and challenges of the emerging sensing technologies for crack detection and characterization.
Emerging Noncontact

Copyright © 2014 John Wiley & Sons, Ltd.


technologies Wireless networks MEMS Sensing skin Self-sensing materials LAE sensing sheet sensing techniques
Advantage Cost reduction of hardware Availability for subsurface High spatial resolution High spatial resolution Low cost, high spatial High spatial
and installation (no connecting crack detection (2D strain-field monitoring), (2D strain-field resolution (2D strain- resolution and
cables), distributed data corrosion detection, wireless monitoring) field monitoring), sensitivity,
acquisition and processing multi-functional, accurate crack
wireless characterization
Challenge Power supply and data Sensor packaging, on-site Immature for widespread Immature for Immature for Real-time crack
transmission losses application (mostly used applications widespread widespread detection (mostly
as accelerometer) and applications applications used as NDT and
high cost not SHM),
complex data
analysis
Direct or Enables direct or indirect Enables direct or indirect Direct Direct Direct Direct
Y. YAO, S.-T. E. TUNG AND B. GLISIC

indirect sensing depending on sensing depending on


sensing type of sensor attached type of sensor
to wireless nodes

DOI: 10.1002/stc
Struct. Control Health Monit. (2014)
CRACK DETECTION AND CHARACTERIZATION TECHNIQUES—AN OVERVIEW

technology. Several field applications were discussed: Imperial Highway bridge in Los Angeles, steel
bridge over I-77 in Mecklenbury Country, and concrete bridge over US-74. The capability of accurate
damage detection from the LiBE system was validated through the aforementioned bridge applications.
Sohn [106] provided an overview of ongoing laser ultrasonic-based SHM researches for damage
detection in civil, mechanical, and aerospace structures. Ultrasonic-based SHM techniques can detect
very small defects even at a remote location on structure. Noncontact laser ultrasonic techniques, in
particular, have better performance than the conventional ultrasonic techniques in three aspects: (1)
they have higher spatial resolution; (2) they are baseline free and have less false positive detections;
and (3) they are faster to deploy. These techniques have been extensively studied in the NDE domain,
and Sohn [106] introduced them in SHM of various structures, such as aircraft, wind turbine, high-
speed train, nuclear power plant, and a real bridge in South Korea. An et al. [107] introduced a laser
ultrasonic-based crack visualization approach for an aluminum plate structure in laboratory, with a
novel image processing technique. The main challenges of noncontact laser ultrasonic techniques come
from the following facts: (1) it requires long time for scanning, more expensive hardware, and complex
data analysis; (2) special surface treatment is needed; and (3) because of the strong laser, eye protection
during testing should be provided.
Another laser technique appearing recently is based on laser thermography. An et al. [108]
presented a noncontact laser lock-in thermography (LLT) technique for fatigue crack detection in a
steel plate structure. The LLT technique dominates conventional laser thermography technique by
the following facts: (1) the laser heat source can be located precisely even at a long distance away;
(2) it is capable to inspect large structure; (3) no special surface treatment is required; and (4) environ-
mental noises can be eliminated. With this technique, the authors proposed a holder exponent filter
approach for crack identification, localization, and quantification. The fatigue crack length in the plate
identified by LLT technique was close to the experimental results, which validated such a method for
damage detection. Besides, this technique has the potential to be applied in real structures, such as
bridges, buildings, train vehicles, and so on. An example of the LLT system and its application for
crack detection is shown in Figure 18.
Advantages and challenges of the emerging sensing technologies are summarized in Table V. The
potential of these emerging technologies for either direct or indirect sensing is also presented in
Table V.
This paper focuses on sensing techniques for crack detection and characterization. However, it is
also important in engineering practice to carry out long-term monitoring of the evolution of existing
and already identified cracks. On the basis of literature review, and the authors’ knowledge and
experience, several of the techniques presented in this paper, such as vibrating-wire-based sensors,
fiber-optic sensor (short-gage, long-gage, and distributed), coaxial cable, and wave-based techniques,
have the capability and have already been applied in real projects for this purpose. All emerging
sensing technologies also have the potential for long-term crack evolution monitoring.

6. CONCLUSIONS
Cracks are critical flaws that affect the behavior and durability of civil structures and infrastructure.
Thus, reliable crack detection and accurate crack characterization are important objectives of SHM.
Numerous crack types and causes are summarized in this paper. Various crack detection and character-
ization techniques are presented, including direct sensing approaches, indirect sensing approaches, and
emerging technologies.
Two major groups of current approaches are identified: direct sensing and indirect sensing. Both
have advantages and challenges. The main advantage of the direct sensing is its reliable crack detection
and, to a certain extent, its accurate crack characterization. This statement is supported with results
reported from real on-site applications, especially with strain sensing. However, the challenge of this
approach is the need for dense arrays of sensors, which requires high hardware, installation, and data
management costs. On the other hand, the main advantage of indirect sensing is its lower cost for hard-
ware and installation. However, the challenge is its reliability in crack detection and accuracy in crack
characterization when applied in real-life settings. The new and emerging sensing technologies, as well

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
Y. YAO, S.-T. E. TUNG AND B. GLISIC

as never-ending developments in computational power, are all paving ways for overcoming the
aforementioned challenges. The authors believe that the improved reliability, robustness, and accuracy
in crack detection and characterization, combined with the low cost that these novel technologies can
offer, have the potential to transform SHM and open the doors for its widespread application, vastly
improving safety and reducing maintenance and life-cycle costs for civil infrastructure.

REFERENCES

1. Chen G, Mu HM, Pommerenke D, Drewniak JL. Damage detection of reinforced concrete beams with novel distributed
crack/strain sensors. Structural Health Monitoring 2004; 3(3):225–243.
2. Sabnavis G, Kirk RG, Kasarda M, Quinn D. Cracked shaft detection and diagnostics: a literature review. The Shock and
Vibration Digest 2004; 36(4):287–296.
3. Subbiah R, Montgomery J, Banks RL. Studies on rotor cracks due to bending and torsional effects. Proceedings of 6th
International Conference on Rotor Dynamics (IFToMM), Sydney, Australia, 2002; 343–349.
4. Bierbaum J, Horst P. Crack propagation in buckling plates: test results and a simplified numerical approach. International
Journal of Structural Integrity 2011; 2(4):373–382.
5. Chen G, McDaniel R, Brower M, Pommerenke D. Crack detectability and durability of coaxial cable sensors in reinforced
concrete bridge applications, Transportation Research Record: Journal of the Transportation Research Board, No. 2172,
Transportation Research Board of the National Academies, Washington, D.C., 2010; 151–156.
th
6. Nilson A, Darwin D, Dolan C. Design of Concrete Structures (14 edn). McGraw-Hill Companies, Inc.: New York, NY,
USA, 2010.
7. Shah SG, Kishen C. Use of acoustic emissions in flexural fatigue crack growth studies on concrete. Engineering Fracture
Mechanics 2012; 87:36–47.
8. Horii H, Shin HC, Pallewatta TM. Mechanism of fatigue crack growth in concrete. Cement & Concrete Composites 1992;
14:83–89.
9. Shah SP, Swartz SE, Ouyang C. Fracture Mechanics of Concrete. John Wiley & Sons, Inc.: New York, NY, USA, 1995.
10. AISC steel construction manual 13th edition, 2010, Part 2, page 2–40, Table II-IV.
11. MacKenzie S. Overview of the Mechanisms of Failure in Heat Treated Steel Components. Failure Analysis of Heat Treated
Steel Components, ASM International: USA, 2008.
12. Head AK. The growth of fatigue cracks. Philosophical Magazine 1953; 33(7):925–938.
13. Ritchie RO. Mechanisms of fatigue-crack propagation in ductile and brittle solids. International Journal of Fracture 1999;
100:55–83.
14. Suratwala T, Wong L, Miller P, Feit MD, Menapace J, Steele R, Davis P, Walmer D. Sub-surface mechanical damage
distributions during grinding of fused silica. Journal of Non-Crystalline Solids 2005; 352(52–54):5601–5617.
15. Varvani-Farahani A, Topper TH. Short fatigue crack characterization and detection using confocal scanning laser microscopy
(CSLM). In Nontraditional Methods of Sensing Stress, Strain, and Damage in Materials and Structures. American Society for
Testing and Materials: West Conshohocken, PA, USA, 1997.
16. Kramarenko O, Kulikovskaya OV. Relationship between depth and length of fatigue crack developing in circular
specimens during bending. Strength of Materials 1978; 10(2):140–146.
17. Blochwitz C. Microcrack propagation in fatigued f.c.c. monocrystals II: crack faces and crack front shapes. Materials
Science and Engineering 1991; 141:49–54.
18. Shine UP. Fatigue failure of structural steel – analysis using fracture mechanics. Proceedings of World Academy of Science:
Engineering and Technology 2008; 48:616–19.
19. Gilbert RI, Ranz G. Time-dependent Behavior of Concrete Structures. Taylor & Francis Group: New York, NY, USA, 2011.
20. St. John D, Poole A, Sims I. Concrete Petrography. John Wiley & Sons, Inc.: New York, NY, USA, 1998.
21. Transportation Research Board. Control of concrete cracking: state of the art. Transportation Research Circular, No.
E-C107, 2006.
22. Concrete Foundation Association. (Available from: http://www.cfawalls.org/foundations/cracking.htm). [accessed 3/15/2013].
23. Gilbert RI. Shrinkage cracking and crack control in fully-restrained reinforced concrete members. In System-Based Vision for
Strategic and Creative Design, Botempi F (ed.). Swets & Zeitlinger B.V., Lisse: The Netherlands, vol. 3, 2003; 1921–1928.
24. Dombrowski KI, Erfurt WGC, Janssen DJ. Identifying D-cracking susceptible aggregates – a comparison of testing
procedures. Proceedings of the International RILEM Workshop on Frost Damage in Concrete, Minnesota. MN, USA:
RILEM Publications S.A.R.L., 1999.
25. Wilson ML, Kosmatka SH. Design and Control of Concrete Mixtures. Portland Cement Association: USA, 2011.
26. Thoft-Christensen P. Corrosion and cracking of reinforced concrete. In Life-Cycle Performance of Deteriorating Struc-
tures, Frangopol D, Brühwiler E, Faber M, Adey B (eds). American Society of Civil Engineers: USA, 2004.
27. Glisic B, Chen J, Hubbell D. Streicker Bridge: a comparison between Bragg-gratings long-gauge strain and temperature
sensors and Brillouin scattering-based distributed strain and temperature sensors. Proceedings of SPIE The International
Society for Optical Engineering, San Diego, CA, USA, 2011a.
28. Glisic B, Yao Y. Fiber optic method for health assessment of pipelines subjected to earthquake-induced ground movement.
Structural Health Monitoring 2012; 11(6):696–711.
29. Widow AL. Strain Gauge Technology (2nd edn). Elsevier Science Publishers, Ltd.: UK, 1992.
30. Heise RE Jr. Low-cycle fatigue-crack indications by strain gages operating in elastic strain fields. Journal of Experimental
Mechanics 1965; 5(7):19A–24A.
31. Glisic B, Inaudi D. Fibre Optic Methods for Structural Health Monitoring. John Wiley & Sons: Chichester, UK, 2007.
32. Inaudi D, Glisic B, Fakra S, Billan J, Redaelli S, Perez JG, Scandale W. Development of a displacement sensor for the
CERN-LHC superconducting cryodipoles. Measurement Science and Technology 2001; 12(7):887–896.
33. Sigurdardottir DH, Glisic B. Neutral axis as damage sensitive feature. Smart Materials and Structures 2013; 22(7):1–18.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
CRACK DETECTION AND CHARACTERIZATION TECHNIQUES—AN OVERVIEW

34. Hubbell D, Glisic B. Detection and characterization of early age thermal cracks in high performance concrete. ACI
Materials Journal 2013; 110(3):323–330.
35. Wipf TJ, Phares BM, Doornink JD, Griemann LF, Wood DL. Evaluation of steel bridges (volume I): monitoring the structural
condition of fracture-critical bridges using fiber optic technology. Final Report, Bridge Engineering Center, Center for
Transportation Research and Education (CTRE), Iowa State University, 2007.
36. Glisic B, Badoux M, Jaccoud J-P, Inaudi D. Monitoring a subterranean structure with the SOFO system. Tunnel Management
International magazine, ITC Ltd, 2000; 2(8):22–27.
37. Glisic B, Inaudi D, Nan C. Pile monitoring with fiber optic sensors during axial compression, pullout, and flexure tests.
Transportation Research Record 2002; 1808:11–20.
38. Posenato D. Model-free data interpretation for continuous monitoring of complex structures. Ph.D. Thesis No. 4481, EPFL,
Lausanne, Switzerland, 2009.
39. Sun S, Pommerenke DJ, Drewniak JL, Chen G, Xue L, Brower MA, Koledintseva MY. A novel TDR-based coaxial cable
sensor for crack/strain sensing in reinforced concrete structures. IEEE Transactions on Instrumentation and Measurement
2009; 58(8):2714–2725.
40. Posey R Jr, Johnson GA, Vohra ST. Strain sensing based on coherent Rayleigh scattering in an optical fibre. ELECTRONICS
LETTERS 2000; 36(20):1688–1689.
41. Glisic B, Inaudi D. Development of method for in-service crack detection based on distributed fiber optic sensors. Structural
Health Monitoring 2012; 11(2):161–171.
42. Inaudi D, Glisic B. Long-range pipeline monitoring by distributed fiber optic sensors. Journal of Pressure Vessel Technology,
Transactions of the ASME 2010; 132(1):0117011–0117019.
43. Ravet F, Briffod F, Glisic B, Nikles M, Inaudi D. Submillimeter crack detection with Brillouin-based fiber-optic sensors.
IEEE Sensors Journal 2009; 9(11): art. no. 5257462, 1391–1396.
44. Klepka A, Staszewski WJ, Jenal RB, Szwedo M, Iwaniec J, Uhl T. Nonlinear acoustics for fatigue crack detection
experimental investigations of vibro-acoustic wave modulations. Structural Health Monitoring 2012; 11(2):197–211.
45. Grosse CU. Acoustic emission localization methods for large structures based on beam forming and array techniques.
Proceedings of Non-Destructive Testing in Civil Engineering, Nantes (NDTCE’09), France, pp. on CD, 2009.
46. Park G, Farrar CR, Lanza di Scalea F, Coccia S. Performance assessment and validation of piezoelectric active sensors in
structural health monitoring. Smart Materials and Structures 2006; 15(6):1673–1683.
47. Maierhofer C, Krankenhagen R, Myrach P, Meinhardt J, Kalisch U, Hennen C, Mecke R, Seidl T, Schiller M. Monitoring
of cracks in historic concrete structures using optical, thermal and acoustical methods. International Conference Built
Heritage 2013, Monitoring Conservation Management, Politecnico di Milano, Italy, on CD, 2013.
48. Shin SW. Elastic Rayleigh Wave for Nondestructive Health Monitoring of Concrete Structure: Theory and Application for
Strength and Crack Monitoring. VDM Verlag: Saarbrücken, Germany, 2008.
49. Rose JL. Ultrasonic Waves in Solid Media. Cambridge University Press: New York, NY, USA, 1999.
50. Towfighi S, Kundu T, Ehsani M. Elastic wave propagation in circumferential direction in anisotropic cylindrical curved
plates. Journal of Applied Mechanics 2002; 69(3):283–291.
51. Sadler DJ, Ahn CH. On-chip eddy current sensor for proximity sensing and crack detection. Sensors and Actuators A:
Physical 2001; 91(3):340–345.
52. Najafi M. Trenchless Technology: Pipeline and Utility Design, Construction, and Renewal. McGraw-Hill: New York City,
USA, 2004.
53. Bachschmid N, Pennacchi P, Tanzi E, Audebert S. Identification of transverse cracks in rotors systems. Proceedings of the
8th International Symposium on Rotating Machinery (ISROMAC-8), Honolulu, Hawaii, 2000a; 1–11.
54. Bachschmid N, Pennacchi P, Tanzi E, Vania A. Identification of transverse crack position and depth in rotor systems.
Meccanica 2000b; 35(6):563–582.
55. Ray S, Kishen JMC. Fatigue crack propagation model and size effect in concrete using dimensional analysis. Mechanics of
Materials 2011; 43(2):75–86.
56. Friswell MI, Penny JET. Crack modeling for structural health monitoring. Structural Health Monitoring 2002; 1(2):139–148.
57. Kim JT, Stubbs N. Crack detection in beam-type structures using frequency data. Journal of Sound and Vibration 2003;
259(1):145–160.
58. Wang JL, Qiao PZ. Improved damage detection for beam-type structures using a uniform load surface. Structural Health
Monitoring 2007; 6(2):99–110.
59. Dimarogonas AD, Papadopoulos CA. Crack detection in turbine rotors. Proceedings of 2nd International Symposium on
Transport Phenomena Dynamics and Design of Rotating Machinery, Honolulu, Hawaii, April 3–6, 1988; 286–298.
60. Dimarogonas AD, Papadopoulos CA. Coupled vibrations of cracked shafts. Journal of Vibration and Acoustics 1992;
114(4):461–467.
61. Papadopoulos CA, Dimarogonas AD. Coupled vibration of cracked shafts. ASME Design Engineering Division 1989;
18(2):7–12.
62. Papadopoulos CA, Dimarogonas AD. Diagnosis of edge cracks in rotating shafts. EPRI, 4th Incipient Failure Detection
Conference, Predictive Maintenance for the 90s, Philadelphia, PA, October 15–17, 1990; 1–26.
63. Miller WH, Brook WR. Shaft Crack Detection Method. 1990; US Patent No. 4,975,855.
64. Miller WH, Brook WR. Crack detection method for operating shaft. 1992; US Patent No. 5,159,563.
65. Brook WR, Miller WH. Crack detection method for shaft at rest. 1991; US Patent No. 5,068,800.
66. Rajab MD, Al-Sabeeh A. Vibrational characteristics of cracked shafts. Journal of Sound and Vibration 1991; 147(3):465–473.
67. Tsai TC, Wang YZ. Vibration analysis and diagnosis of a cracked shaft. Journal of Sound and Vibration 1996; 192(3):607–620.
68. Pai PF, Young LG, Lee SY. A dynamics-based method for crack detection and estimation. Structural Health Monitoring
2003; 2(1):5–25.
69. Casciati F, Domaneschi M, Inaudi D, Figini A, Glisic B, Gupta A. Long-gauge fibre-optic sensors: a new approach to dynamic
system. Proceedings of the 3rd European Conference on Structural Control (3ECSC), Vienna, Austria, 1, 2004; M3-5.
70. Doebling SW, Farrar CR, Prime MB. A summary review of vibration-based damage identification methods. The Shock and
Vibration Digest 1998; 30(2):91–105.
71. Kanjilal PP, Palit S. Modeling and prediction of time series using singular value decomposition and neural networks.
Computers and Electrical Engineering 1995; 21(5):299–309.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
Y. YAO, S.-T. E. TUNG AND B. GLISIC

72. Zhao M, Luo ZH. An expert system of crack monitoring and diagnosing for rotating machines. Proceedings of Conference
on Rotating Machine Dynamics, Venice, Italy, 1992; 84–91.
73. Moyo P, Brownjohn JMW. Detection of anomalous structural behaviour using wavelet analysis. Mechanical Systems and
Signal Processing 2002; 16(2-3):429–445.
74. Kiremidjian AS, Sarabandi P, Cheung A, Cabrera C, Nair KK, Kiremidjian G. Algorithm for identification of damage on
bridge piers. Proceedings of the International. Symposium on Life-Cycle Civil Engineering, Varenna, Italy, 2008; 935–940.
75. Posenato D, Lanat F, Inaudi D, Smith IFC. Model-free data interpretation for continuous monitoring of complex structures.
Advanced Engineering Informatics 2008; 22(1):135–144.
76. Adewusi SA, Al-Bedoor BO. Detection of propagating cracks in rotors using neural networks. American Society of
Mechanical Engineers, Pressure Vessels and Piping Division 2002; 447:71–78.
77. Yin T, Lam HF, Chow HM. A Bayesian probabilistic approach for crack characterization in plate structures. Computer-
Aided Civil and Infrastructure Engineering 2010; 25(5):375–386.
78. Sbarufatti C, Manes A, Giglio M. ANN based Bayesian hierarchical model for crack detection and localization over
helicopter fuselage panels. In Advances in Safety, Reliability and Risk Management, CRC Press: Taylor & Francis Group,
London, Chapter 47, 2012; 378–385.
79. Umesha PK, Ravichandran R, Sivasubramanian K. Crack detection and quantification in beams using wavelets. Computer-
Aided Civil and Infrastructure Engineering 2009; 24(8):593–607.
80. Feldman M, Seibold S. Damage diagnosis of rotors: application of Hilbert transform and multi-hypothesis testing. Reports
of the Institut für Techno- und Wirtschaftsmathematik (ITWM), No. 2., 1998; 1–23.
81. Rehman AU, Worden K, Rongong JA. Damage diagnosis of rotors: application of Hilbert transform and multi-hypothesis
testing. Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science 2012;
226(11):2610–2626.
82. Vakil-Baghmisheh MT, Peimani M, Sadeghi MH, Ettefagh MM. Crack detection in beam-like structures using genetic
algorithms. Applied Soft Computing 2008; 8(2):1150–1160.
83. Lynch JP, Loh KJ. A summary review of wireless sensors and sensor networks for structural health monitoring. The Shock
and Vibration Digest 2006; 38(2):91–128.
84. Spencer Jr BF, Yun CB. Wireless sensor advances and applications for civil infrastructure monitoring. NSEL Report No.
024, University of Illinois at Urbana-Champaign, 2010.
85. Wang KCP, Li Q. Applicability of microelectronic and mechanical systems (MEMS) for transportation infrastructure man-
agement. Final Report for Project MBTC-2056, Dept. of Civ. Eng., U. of Arkansas, Fayetteville, 2008.
86. Pozzi M, Zonta D, Trapani D, Athanasopoulos N, Amditis AJ, Bimpas M, Garetsos A, Stratakos YE, Ulieru D. MEMS-based
sensors for post-earthquake damage assessment. Journal of Physics: Conference Series 2011; 305(1): art. no. 012100.
87. Albarbar A, Badri A, Sinha JK, Starr A. Performance evaluation of MEMS accelerometers. Journal of the International
Measurement Confederation 2009; 42(5):790–795.
88. Zhang Y. In situ fatigue crack detection using piezoelectric paint sensor. Journal of Intelligent Material Systems and
Structures 2006; 17(10):843–852.
89. Loh KJ, Hou TC, Lynch JP, Kotov NA. Nanotube-based sensing skins for crack detection and impact monitoring of structures.
Proceedings of the 6th International Workshop on Structural Health Monitoring, Stanford, CA, pp. on CD, 2007a.
90. Schumacher T, Thostenson ET. Development of structural carbon nanotube-based sensing composites for concrete
structures. Journal of Intelligent Material Systems and Structures 2013. doi:10.1177/1045389X13505252.
91. Zonta D, Chiappini A, Chiasera A, Ferrari M, Pozzi M, Battisti L, Benedetti M. Photonic crystals for monitoring fatigue
phenomena in steel structures. Proceedings of SPIE, San Diego, CA, USA, 2009.
92. Loh KJ, Hou TC, Lynch JP, Kotov NA. Carbon nanotube sensing skins for spatial strain and impact damage identification.
Journal of Nondestructive Evaluation 2009; 28:9–25.
93. Loh KJ, Kim J, Lynch JP, Kam NWS, Kotov NA. Multifunctional layer-by-layer carbon nanotube-polyelectrolyte thin
films for strain and corrosion sensing. Smart Materials and Structures 2007b; 16(2):429–438.
94. Pyo S, Loh KJ, Hou TC, Jarva E, Lynch JP. A wireless impedance analyzer for automated tomographic mapping of a
nanoengineered sensing skin. Smart Structures and Systems 2011; 8(1):139–155.
95. Pour-Ghaz M, Weiss J. Application of frequency selective circuits for crack detection in concrete elements. Journal of
ASTM International 2011; 8(10):11.
96. Chung DDL. Multifunctional Cement-Based Materials. Marcel-Dekker: New York, NY, USA, 2003.
97. Lynch JP, Hou TC. Conductivity-based strain and damage monitoring of cementitious structural components. Proceedings
of SPIE 2005; 5765:419–429.
98. Glisic B, Verma N. Very dense arrays of sensors for SHM based on large area electronics. Structural Health Monitoring
2011: Condition-Based Maintenance and Intelligent Structures - Proceedings of the 8th International Workshop on
Structural Health Monitoring 2, 2011; 1409–1416.
99. Arias AC, MacKenzie JD, McCulloch I, Rivnay J, Salleo A. Materials and applications for large area electronics: solution-
based approaches. Chemical Reviews 2010; 110(1):3–24.
100. Someya T, Pal B, Huang J, Katz HE. Organic semiconductor devices with enhanced field and environmental responses for
novel applications. MRS Bulletin 2008; 33(7):690–696.
101. Lecompte D, Vantomme J, Sol H. Crack detection in a concrete beam using two different camera techniques. Structural
Health Monitoring 2006; 5(1):59–68.
102. Jahanshahi MR, Masri SF. Nondestructive vision-based approaches for condition assessment of structures. Proceedings of
SPIE, San Diego, CA, USA, 2011.
103. Uhl T, Kohut P, Holak K, Krupinski K. Vision based condition assessment of structures. Journal of Physics: Conference
Series 2011; 305(1):1–10.
104. Liu WQ, Chen SE, Sajedi A, Hauser E. The role of terrestrial 3D LiDAR scan in bridge health monitoring. Proceedings of
SPIE, San Diego, CA, USA, 2010.
105. Liu WQ, Chen SE. Reliability analysis of bridge evaluations based on 3D light detection and ranging data. Structural
Control and Health Monitoring 2013; 20:1397–1409.
106. Sohn H. Laser based structural health monitoring for civil, mechanical and aerospace systems. Proceedings of SPIE, San
Diego, CA, USA, 2012.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
CRACK DETECTION AND CHARACTERIZATION TECHNIQUES—AN OVERVIEW

107. An YK, Park B, Sohn H. Complete noncontact laser ultrasonic imaging for automated crack visualization in a plate. Smart
Materials and Structures 2013; 22(2):1–10.
108. An YK, Kim JM, Sohn H. Laser lock-in thermography for fatigue crack detection in an uncoated metallic structure.
Proceedings of SPIE, San Diego, CA, USA, 2013.
109. Inspectapedia 2013, (Available from: http://inspectapedia.com/structure/SlabCracks4.htm). [last accessed on 04/24/2013].
110. Wikipedia 2013, (Available from: http://en.wikipedia.org/wiki/Alkali–silica_reaction). [last accessed on 04/24/2013].
111. Metallurgical Technologies 2013, (Available from: http://www.met-tech.com/preheater-tube-failure.html). [last accessed
on 04/24/2013].
112. Measurement specialties 2013, (Available from: http://www.meas-spec.com/product/t_product.aspx?id=5435), [last
accessed on 04/23/2013].

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc

You might also like